A Magnetic Resonance Imaging (hereinafter, MRI) apparatus measures a Nuclear Magnetic Resonance (hereinafter, NMR) signal generated by atomic nuclear spin comprising an object, in particular human body tissue and images shapes and tissue of the head, abdomen, extremities, etc. two-dimensionally or three-dimensionally.
In case of obtaining an image using an MRI apparatus, an image with various tissue contrasts can be obtained by changing parameters such as an Echo Time (hereinafter, described as TE) and a Repetition Time (hereinafter, described as TR) or performing image computation. A method to obtain an image separating water and fat for image accuracy improvement is referred to as the Dixon method.
Additionally, there are the 2-point Dixon method and the 3-point Dixon method with static magnetic field correction in which a function to correct influence of static magnetic field inhomogeneity is added to the Dixon method. When a phase rotation amount due to the static magnetic field inhomogeneity is calculated using these methods, arithmetic processing referred to as a phase unwrapping process is performed to prevent water and fat from exchanging. The phase unwrapping process fixes discontinuities caused because a phase beyond the range of −π to π (this state is referred to as “rotating around a principal value”) is expressed within the range of −π to π again and makes a spatial phase change continuous to express with a value of the phase beyond the range of −π to π.
When rotating around a principal value, two main causes were found in examination by the inventors of the present invention: a case where phase differences between the adjacent pixels is large due to large static magnetic field inhomogeneity and a case where a Signal Noise Ratio (hereinafter, SNR) is low.
The above unwrapping processing is performed for two-dimensional or three-dimensional data, and the regional expansion method is used. The regional expansion method determines a pixel to start processing first (hereinafter, described as a staring pixel) and expands the processing from the starting pixel to the adjacent pixels spatially, and a pixel value that has already been processed is utilized in processing the adjacent pixels.
In the patent literature 1, a method to reduce principal value rotation by processing from a pixel with a smaller phase difference in order from among the adjacent pixels has been suggested. Specifically, a phase difference inclination image is generated, and a pixel with a smaller phase difference is processed first.